updated 15 September 2019

OSMuf is a project to explore the use of Python and OpenStreetMap for quantifying urban form, specifically the size and shape of urban/city blocks and built density.

It is in a large part inspired by the work of Geoff Boeing on OSMnx, Berghauser Pont and Haupt on the 'Spacemate', Barthélemy and Louf on Measuring Urban Form, and Philip Steadman on 'integrating 'Spacemate' with the work of Martin and March'.

Large scale studies of urban form often approximate city blocks by generating polygons from the centrelines of the transport network. The approach of OSMuf is to instead work with urban blocks as objects in themselves with their own spatial definition. Loosely this could be categorised as the difference between 'gross urban blocks' defined by transport network centrelines and 'net urban blocks' defined by the line separating privately owned land or land dedicated to public amenity from land dedicated to the transport network.

There are benefits to dealing with urban blocks directly rather than just as the resultant geometry from the transport network. First, as others have noted, when polygonising the transport network cul-de-sacs tend to disappear - this is misleading especially when dealing with perimeter blocks because length of accessible frontage has a much greater impact on developable area than the absolute area of the urban block. Second, centrelines on their own say little about street width and, as many studies have emphasised, the resultant angle from the street width and building height is often used as a proxy for available daylight within buildings. Third, the limits of urban blocks are generally observable in the real world - this is important for being able to measure density against data available freely in OpenStreetMap which is a topographic map not a cadastre.

Urban Form and Density are recognised as having an impact on quality of life in urban environments. In terms of urban form, towers in otherwise empty estates are blamed for destroying the streetlife associated with traditional perimeter blocks. In terms of density, excessive residential densities are blamed for poor quality of life with limited access to green space yet low residential densities can make provision of public services and amenities unfeasibly expensive.

This project aims to establish a method for quantifying existing built environments based on data from, or added to, OpenStreetMap and using an open source Python library based primarily on osmnx, geopandas and matplotlib. Ultimately it aims to put some relatively simple measures against a variety of urban environments to contribute to an informed discussion about how best to design sustainable towns and cities.

Development of OSMuf has been carried out in a Python 3.7 Miniconda environment on Ubuntu and the installation instructions below reflect this. It has also been installed into a Miniconda environment on Windows. On Windows problems were encountered with missing dll files. This was resolved by installing certain libraries, including gdal, from pip. See here for further instructions. It has also been installed and run on a Raspberry Pi - please see the separate README_RaspberryPi.md.

Set up of Miniconda followed Ted Petrou's guidance.

Create and activate a new empty Miniconda environment

conda create -n osmuf_test
conda activate osmuf_test

add the conda-forge channel to this environment and set to strict priority (i.e. always prefer packages from conda-forge)

conda config --env --add channels conda-forge
conda config --env --set channel_priority strict

check the channels and priority with

conda config --show channels
conda config --show channel_priority

OSMnx can then be installed with

conda install osmnx

Jupyter has to be installed into the environment otherwise it may not find osmuf

conda install jupyterlab

OSMuf should install seaborn itself, otherwise install it with

conda install seaborn

Either download and expand the zip of OSMuf from its github repository or clone it using

git clone https://github.com/AtelierLibre/osmuf.git

Change directory into the downloaded osmuf folder

cd osmuf/

Install OSMuf with pip

pip install .

or

python setup.py install

create a folder for the final images

mkdir local_images

Start jupyter lab (jupyter notebook should work as well)

jupyter lab

Check that jupyter is running using the kernel from the new environment by running the following code in a new cell of a notebook

import sys
sys.executable

This should list the python kernel as running from something like

../anaconda3/envs/osmuf_test/bin/python

If this isn't the case refer to the Ted Petrou link above.

Click on the OSMuf_v0.1.ipynb notebook to start it.

Run each of the cells in turn to generate the visualisations for the four sample areas.

To change the sample area, change the place name in cell [4] to match one of the other names in the dictionary in cell [1].